arXiv:1111.2471v1 [math.DG] 10 Nov 2011
FINITE VOLUME
NADINE GROSSE
Abstract. We prove the existence of a solution of the Yamabe equation on complete manifolds with finite volume and positive Yamabe invariant. In order to circumvent the standard methods on closed manifolds which heavily rely on global (compact) Sobolev embeddings we approximate the solution by eigenfunctions of certain conformal complete metrics.
This also gives rise to a new proof of the well-known result for closed manifolds and positive Yamabe invariant.
1. Introduction
Yamabe examined whether a closed n-dimensional Riemannian manifold (M, g) (n ≥ 3) possesses a metricgconformal togwith constant scalar curvature. His striking idea was the consideration of the so-called Yamabe invariant, see Definition 1, which gave the possibility to view the question as a variational problem. Works from Aubin [1], Schoen [10] and Trudinger [14] answered the question of Yamabe affirmatively.
There are several possibilities to generalize the Yamabe problem to open (i.e. noncompact and without boundary) manifolds.
One possibility is simply to pose the same question as Yamabe did. On open manifolds, this gives much more freedom. We want to make this more precise by comparing to the closed case:
On closed manifolds, if the Yamabe invariantQis nonpositive, then every conformal metric having constant scalar curvaturec and unit volume fulfillsc =Q. In case thatQ <0 this conformal metric is even unique. If Q > 0 and there is a conformal metric with constant scalar curvaturec, one see immediately that c≥Q. But nevertheless, on closed manifolds in all cases a conformal metric with constant scalar curvature has the same sign as the corresponding Yamabe invariant.
On open manifolds, this is no longer true, an easy example is given by an open ball in the Euclidean space. Its Yamabe invariant is the one of the standard sphere, but it carries conformal metrics of constant scalar curvature of all signs: The original Euclidean metric has zero scalar curvature, the spherical metric has constant positive and the hyperbolic metric has constant negative scalar curvature. But all those metrics are conformally equivalent.
That’s why the question is often posed more restrictively. A first possibility is to fix the sign of the constant scalar curvature and/or ask additionally for completeness. This was done by many authors and many results with positive and negative answers were obtained, see for example [2], [7], [15].
2000Mathematics Subject Classification. 53C21, 53A30, 35R01.
Key words and phrases. Yamabe problem, scalar curvature, complete manifolds of finite volume.
I like to thank the Hausdorffcenter in Bonn where parts of this paper were written down for its hospitality.
1
The second is to stick to the original Yamabe problem and ask for solutions to the Euler- Lagrange equation of the Yamabe problem, i.e. unit volume metrics with constant scalar curvature Q. This version was studied for example for manifolds with bounded geometry and positive scalar curvature in [8] using a compact exhaustion of the open manifold and for manifolds bounded geometry and positive Yamabe invariant in [5] using weighted Sobolev embeddings.
In this paper we consider the second type of a noncompact Yamabe problem on complete manifolds of finite volume.
Let (Mn, g) be ann-dimensional complete connected Riemannian manifold of finite volume and n ≥3. Let Lg =an∆g+ scalg be the conformal Laplacian where scalg is the scalar curvature of the metric gandan = 4n−1n−2.
Definition 1. The Yamabe invariant of(M, g)is given by
Q(M, g) = inf (
Qg(v) :=
R
MvLgvdvolg
kvk2Lp(g)
v∈Cc∞(M), v6= 0 )
where p= n−22n andCc∞(M)denotes the set of compactly supported real valued functions on M.
Q is conformally invariant which is seen from the conformal transformation formula of the conformal Laplacian: For g=f2g wheref ∈C>0∞(M) is a smooth positive real function on M we have
Lgv=f Lgv wherev=f−n−22v.
The Yamabe invariant is given as a variational problem. Its Euler-Lagrange equation is Lgv=Qvp−1 v∈H12, kvkLp(g)= 1. (1) The aim of this paper is to study the existence of a smooth positive solution of (1) for complete manifolds of finite volume.
The standard proof for the Yamabe problem on closed manifolds heavily relies on the exis- tence of compact Sobolev embeddings. On complete open manifolds of finite volume there do not even exist continuous Sobolev embeddings H12֒→Lp, [6, Lem. 3.2]. That’s why we will use a different approach by approximating the desired solution by certain eigenfunctions of conformal metrics, cp. Section 3.
In the standard proof on closed manifolds one uses the subcritical Yamabe problem to get solutions of differential equations that are somehow ’near’ to the desired Euler-Lagrange equations. This allows to show converges of a sequence of those solutions which then serve as test functions for the critical problem. In our approach here the eigenfunctions will play the role of these subcritical solutions and we obtain
Theorem 2. Let(M, g)be an open complete manifold of finite volume with 0< Q(M, g)<
Q(M, g)andk(scalg)−kn2 <∞where (scalg)−:=−min{scal,0}.
Then, there exists a smooth positive solution v∈H12 of Lgv=Qvp−1 withkvkLp(g)= 1.
Q(M, g) is the Yamabe invariant at infinity, see Definition 4, and replacesQ(Sn) that appears at this point in the closed case, cf. Remark 5.
The non-existence of a continuous Sobolev embeddingH12֒→Lphas the following straight- forward implications: If Q > 0, scalg cannot be bounded from above. Moreover, if v is a solution as in Theorem 2, g =vn−2n2g is a metric with finite volume and constant scalar curvature and for all v ∈ Cc∞(M) kvkLp(g) ≤(max{an, Q})12kvkH21(g). Thus,g cannot be complete.
The method used to prove Theorem 2 also gives rise to a different proof for the closed case with positive Yamabe invariant, see Theorem 14. Moreover, the method can be adapted to similar contexts, e.g one can obtain similar results for the spinorial Yamabe invariant, cf.
[4].
2. Preliminaries In this section we collect some facts on the Yamabe invariant.
Remark 3. On complete manifolds instead of taking the infimum overCc∞in the definition of the Yamabe invariant 1 one could as well take the infimum overv ∈ L2∩H1,loc2 with
R
MvLgvdvolg
<∞. This is seen when consideringQg(ηiv) for suitable cut-off functions ηi withηi→1.
Definition 4. [9]Let(M, g)be an openn-dimensional manifold with a compact exhaustion Ki fulfilling Ki ⊂ Ki+1 M and ∪iKi = M. Then the Yamabe invariant at infinity is defined as
Q(M, g) := lim
i→∞Q(M\Ki, g).
Note that Q(M \Ki, g) ≤ Q(M \Ki+1, g) since when considering only a subset less test functions can be used in Definition 1. Together withQ(M, g)≤Q(Sn) [11] whereQ(Sn) = n(n−1)vol(Sn)2nis the Yamabe invariant of the sphere with the standard metric the sequence Q(M \Ki, g) is monotonically increasing and bounded. Thus,Qalways exists and it holds Q(M, g)≤Q(Sn). Furthermore,Qdoes not depend on the choice of the sequenceKi. Remark 5. We note that the conditionQ(M, g)< Q(Sn) in Theorem 2 replacesQ(M, g)<
Q(Sn) that appears in the closed case. This can be seen since forp∈M we haveQ(M, g) = Q(M\ {p}, g) [11, Lem. 2.1] andQ(M\ {p}, g) = limi→∞Q(Bǫ(p), g) =Q(Sn) whereBǫ(p) is a ball aroundpwith radiusǫ.
The blow-up argument in the standard proof of the Yamabe problem [12] which rules out concentration phenomena at a fixed point shows that for fixed x ∈ M Q(Bǫ(x), g) → Q(Rn, gE) =Q(Sn, gst) asǫ→0. We will need the following slight generalization:
Lemma 6. For all compact subsets U ⊂M andδ >0 there is anǫ=ǫ(U, δ)>0 such that for allx∈U: Q(Bǫ(x), g)≥Q(Sn)−δ.
Proof. LetU andδbe fixed. Then for eachx∈U letǫ(x) be the maximal radius such that Q(Bǫ(x), g)≥Q(Sn)−δ is fulfilled. Setǫ= infX∈Uǫ(x). Suppose ǫ= 0. Then there is a sequence xi ∈ U withǫ(xi) →0. Since U is compact, xi → x ∈U. Note that on closed manifolds Q depends smoothly on g in the C2-topology [3, Proof of Prop. 7.2.]. Thus, ǫ(xi)→ǫ(x)>0 which is a contradiction. Thus,ǫ >0.
3. Nonnegative Yamabe invariants and the L2-spectrum On closed manifolds and ifQ≥0,
Q(M, g) = inf{µ(Lg)|g∈[g]}
whereµ(Lg) is the lowest eigenvalue of the conformal LaplacianLgand [g] :={g=f2g|f ∈ C>0∞(M)} denotes the conformal class ofg.
On general manifolds the spectrum ofLg does not only contain eigenvalues but there can be residual and continuous spectrum. Moreover, in generalLg is even not essentially self- adjoint.
We consider
µ(Lg) = inf ( R
MvLgvdvolg
kvk2L2(g)
v∈Cc∞(M) )
.
IfLg is essentially self-adjoint,µ(Lg) is the minimum of the spectrum ofLg. Remark 7. IfQ≥0 and vol(M, g) = 1, thenR
MvLgvdvolg ≥Qkvk2p≥Qkvk22, i.e. µ(Lg)≥ QandLgis bounded from below. Then,Lgis essentially self-adjoint onCc∞(M), [13, Thm.
1.1] and possesses only eigenvalues and essential spectrum. Moreover, the spectrum is real.
If it is clear from the context to which Riemannian manifold (M, g) we refer, we abbreviate k.ks:=k.kLs(g).
Lemma 8. Let (M, g)be a Riemannian manifold withQ≥0. Then Q(M, g) = inf{µ(Lg)| g∈[g],vol(M, g) = 1}.
If (M, g)has additionally unit volume,
Q(M, g) = inf{µ(Lg) |g=f2g,vol(M, g) = 1, ∃ a compact subset Kf⊂M : f|M\Kf = 1}.
If for a function f ∈C>0∞(M) such a compact subsetKf exists, we shortly say that f ≡1 near infinity. The proof of the first part is the same as in the closed case. But since we are not aware of a reference we shortly give the proof.
Proof. Without loss of generality we can assume that g already has unit volume. Since Q≥0,R
MvLgvdvolg≥0 for allv∈Cc∞(M).
From Remark 7 we haveµ(Lg)≥Q(M, g) for all conformal metricsg∈[g] with unit volume.
On the other hand, let vi ∈ Cc∞(M) be a minimizing sequence for Q with kvikp = 1 and R
MviLgvidvolg → Q. Set gi = kvi+i−1k−1p (vi+i−1)n−24
g. Then, vol(M, gi) = R
M kvi+i−1k−1p (vi+i−1)p
dvolg = 1 (Note that kvi+i−1kp ≤ kvikp+i−1 = 1 +i−1 is finite.). Moreover,
kvik2L2(gi)= Z
M
kvi+i−1k−1p (vi+i−1)n−42
vi2dvolg
≥ Z
M
(1 +i−1)−n−24 (vi+i−1)n−24 v2idvolg
≥(1 +i−1)−n−24 Z
M
vipdvolg= (1 +i−1)−n−24 . Hence,
0≤µ(Lgi)≤ R
MviLgividvolgi
kvik2L2(gi)
≤ R
MviLgvidvolg
(1 +i−1)−n−42 →Q(M, g) as i→ ∞which finishes the proof of the first claim.
Let now (M, g) be complete and of finite volume andvibe the test sequence of above. LetKi
be a sequence of compact subsets with supp vi⊂Ki. The valuekvik−2L2(gi)
R
MviLgividvolgi
only depends on the metricgi on suppvi. Thus, we can deformgisuch that the conformal factor fi = 1 outside a compact subset Kfi with Ki ⊂⊂ Kfi M, fi2g ≡ gi on Ki and vol(fi2g) = 1.
In particular, if g was complete, all thoseg are also complete.
Next we study the Yamabe invariant if essential spectrum is present.
Lemma 9. Let (M, g) be a complete Riemannian manifold of unit volume. Let Lg be essentially self-adjoint onCc∞(M)and let the essential spectrum ofLgbe non-empty. Then Q(M, g) =Q(M, g)≤0.
Proof. Letµbe in the essential spectrum ofLg. Then there is a sequencevi ∈Cc∞(M\Bi) where Bi a ball with radius i around a fixed point z ∈ M such that k(Lg−µ)vik2 → 0, kvik2= 1 andvi→0 weakly in L2. Then, using 1 =kvik2≤ kvikpvol(M \Bi)n2 and, thus, kvikp≥1 we estimate
Q(M \Bi)≤ R
MviLgvidvolg
kvik2p ≤kLgvik2kvik2
kvik2p ≤ (k(Lg−µ)vik2+|µ|)kvik2
kvik2p
≤ (k(Lg−µ)vik2+|µ|)kvikpvol(M\Bi)n2 kvik2p
≤(k(Lg−µ)vik2+|µ|)vol(M\Bi)n2
where the right hand-side goes to zero asi→ ∞.
From that and Remark 7 it follows directly
Corollary 10. If(M, g)is a complete Riemannian manifold of unit volume withQ(M, g)>
0 or Q(M, g) > Q(M, g) = 0, there exists a sequence of gi = fi2g with fi ∈ C>0∞(M), R
Mfindvolg= 1 and eigenvaluesµi:=µ(Lgi)→Q asi→ ∞.
4. Proof of Theorem 2
From Corollary 10 we have: If Q > 0, there exists a sequence of eigenfunctions vi with Lgivi = µivi and R
M|vi|2dvolgi = 1. vi is eigenfunction to the lowest eigenvalue µi of gi = fi2g (fi = 1 near infinity) and, hence, positive. Viewing these equation w.r.t. the reference metricg we obtain the following setting
Lgvi=µifi2vi
Z
M
fi2v2idvolg = 1, Z
M
findvolg= 1, µiցQ, vi>0.
Firstly we note thatR
M|vi|2dvolgi =R
Mfi2vi2dvolgi= 1 andfi= 1 outside a compact subset impliesvi∈L2(g).
Moreover, due to Remark 3 vi can serve as a test function for Q and, thus, Qkvik2p ≤ R
MviLgvidvolg = µi. Since Q > 0 vi ∈Lp(g) and if then i → ∞, we obtain kvikp →1.
Thus,vi is uniformly bounded inLp(g) and, due to the finite volume, also inL2(g).
From
µi= Z
M
viLgvidvolg =ankdvik22+ Z
M
scalgv2idvolg
≥ankdvik22− Z
M
(scalg)−vi2dvolg≥ankdvik22− k(scalg)−kn2kvik2p
and the assumption thatk(scalg)−kn2 <∞ we see thatkdvik2 is also uniformly bounded.
Summarizingvi is uniformly bounded in H12 and, hence,vi →v ≥0 weakly in H12 and in Lp. Moreover,R
Mfindvolg = 1 implies that there is f ∈ Ln such thatfi2 →f2 weakly in Ln2.
Lemma 11. Let fi2→f2 weakly inLn2 and vi→v weakly inH12.
i) Then fi2vi→f2v inLs(U)for all compact subsetsU ⊂M and1< s < q= n+22n . ii) If additionally Lgvi =µifi2vi andµi→Q,f andv weakly fulfill Lgv=Qf2v.
Proof. i) We fixw∈Ls∗(U) with 1s+s1∗ = 1. Then
Z
U
(fi2vi−f2v)wdvolg
≤ Z
U
|fi2−f2| |vw|dvolg+ Z
U
fi2|vi−v| |w|dvolg
The weak convergence vi →v in H12 implies strong convergence vi → v on Lp′(U) for all 1≤p′ < p. We chooseq′ such thatp > q′> n−2n . Then H¨older inequality implies
kvwkLn−n2(U)≤ kvkLq′(U)kwk
L
nq′
(n−2)q′ −n(U)
<∞
if (n−2)qnq′′−n ≤s∗. The choice ofq′ impliesp < (n−2)qnq′′−n <∞ and, thus,p < s∗<∞ and 1< s < q.
Hence,R
U|fi2−f2| |vw|dvolg →0 asi→ ∞. For the second summand of the above inequality we have
Z
U
fi2|vi−v| |w|dvolg≤ kvi−vkLq′(U)kfi2wk
L
q′
q′ −1(U)
≤ kvi−vkLq′(U)kfik2Ln(U)kwk
L
nq′
(n−2)q′ −n(U)
≤ kvi−vkLq′(U)kwkLs(U)
→0 asi→ ∞.
ii) Letw∈Cc∞(M).
Z
M
(Lgv−Qf2v)wdvolg
= Z
M
(Lgv−Lgvi+µifi2vi−Qf2v)wdvolg
≤an
Z
M
(dv−dvi)dwdvolg
+ Z
M
(v−vi)(scalgw)dvolg
+µi
Z
M
(fi2vi−f2v)wdvolg
+|Q−µi| Z
M
f2|vw|dvolg
All summands on the right-hand side tend to zero as i → ∞ sincevi → v weakly in H12 (note that scalgw∈Cc∞⊂L2), part i) andµi →Q.
In order to finish the proof of Theorem 2 it remains to show that f2=vp−2 andkvkp= 1.
We start with a non-vanishing result.
Lemma 12. In the setting of Lemma 11 and assuming 0< Q(M, g)< Q(Sn),v does not vanish identically.
Proof. We prove by contradiction and assume v ≡ 0 and, hence, R
Uvsidvolg → 0 for all compact subsetsU ⊂M and 1≤s < p.
Firstly, we want to show that then alsoR
Uvipdvolg→0 for all compact subsetsU ⊂M. For that, we assume the contrary, i.e. kvikLp(U) > C(U)>0 and consider small balls B2ǫ(x) with x∈M. We chooseǫsmall enough such that for allx∈U Q(B2ǫ(x))> Q(M, g). Due to Q(Sn) > Q(M, g) and Lemma 6 this is always possible. Then we cover U by finitely many of those ballsB2ǫ(x) and define smooth cut-off functionsηǫ,x compactly supported in B2ǫ(x) that are 1 onBǫ(x) and|dηǫ,x| ≤2ǫ−1. Then we estimate
Q(B2ǫ(x), g)≤ R
B2ǫηǫ,xviLg(ηǫ,xvi)dvolg
R
B2ǫ(x)ηpǫ,xvpidvolg
p2 = R
B2ǫηǫ,x2 viLgvidvolg+anR
B2ǫ|dηǫ,x|2v2idvolg
R
B2ǫ(x)ηpǫ,xvipdvolg
2p
≤µi
R
B2ǫ(x)ηǫ,x2 fi2v2idvolg+an4 ǫ2
R
B2ǫ(x)v2i R
B2ǫ(x)ηǫ,xp vpidvolg
2p ≤µi+an
4 ǫ2
R
B2ǫ(x)v2idvolg
R
Bǫ(x)vipdvolg
2p
≤µi+an
4 ǫ2C(U)2
Z
B2ǫ(x)
v2idvolg
where in the second last step we used the H¨older inequality to estimate the summand including µi. If i tends to ∞, we obtain Q(B2ǫ(x), g) ≤ Q which is a contradiction to Q(B2ǫ(x))> Q. Thus,kvikLp(U)→0 asi→ ∞.
Next, letχR be a smooth cut-off function withχR= 0 on BR:=BR(z) for a fixedz∈M, χR= 1 onM \B2R and|dχR| ≤2R−1. Then
Q= lim
i→∞
Z
M
viLgvidvolg= lim
i→∞
Z
M
χ2RviLgvidvolg+µi
Z
M
(1−χ2R)fi2v2idvolg
≥ lim
i→∞
Z
M
χ2RviLgvidvolg+µi
Z
BR
fi2vi2dvolg
≥ lim
i→∞
Z
M
χRviLg(χRvi)dvolg−an
Z
M
|dχR|2vi2dvolg+µi
Z
BR
fi2vi2dvolg
≥ lim
i→∞
Q(M \BR, g)kχRvik2p−4an
R2kvik2L2(B2R)+µi
Z
BR
fi2v2idvolg
≥ lim
i→∞
Q(M \BR, g)(kvikp− k(1−χR)vikp)2−4an
R2kvik2L2(B2R)+µi
Z
BR
fi2v2idvolg
≥ lim
i→∞
Q(M \BR, g) kvikp− kvikLp(B2R)2
−4an
R2kvik2L2(B2R)+µi
Z
BR
fi2vi2dvolg
WithkvikLs(U)→0 for 1≤s≤pon compact subsets U⊂M,kvikp→1 and Z
BR
fi2vi2dvolg≤ kfiknkvik2Lp(BR)≤ kvik2Lp(BR)→0 we obtain for allRthat
Q(M, g)≥Q(M\BR, g).
That contradictsQ > Q. Thus,v6= 0.
Now we can estimate
Q≤ R
MvLgvdvolg
kvk2p
≤Q R
Mf2v2dvolg
kvk2p
≤Qkfk2nkvk2p kvk2p
≤Q.
Hence, there is already equality. In particular, from the equality case in the used H¨older inequality we getf2=vp−2and 1 =kfkn=kvkp. Smoothness ofvis obtained by standard local elliptic regularity theory. By the maximum principle one sees that v is everywhere
positive which concludes the proof of Theorem 2.
Standard local elliptic regularity also gives thatv is locally inC2,α.
Remark 13 (On the assumption on the scalar curvature).
In Theorem 2 we assume thatk(scalg)−kLn2(g)<∞. If the Yamabe invariantQ(g) =−∞, this could never be true. But in general it can happen that even thoughk(scalg)−kLn2(g)=
∞, Q is finite and even positive. The easiest example is the standard hyperbolic space Hn which has constant negative scalar curvature, infinite volume but the Yamabe invariant of the standard sphere. From this we can even easily construct an example with finite volume: Firstly, we note thatk(scalg)−kLn2(g)is scale invariant. Let us take a ballB in the hyperbolic space with k(scalg)−kLn2(g) = 1 and then rescale it such that the rescaled ball Bi has volume i−2. If we consider the disjoint sum of the Bi, we obtain an example for a (disconnected) Riemannian manifold of finite volume andk(scalg)−kLn2(g)=∞.
We assume that Q(g) = −∞ if and only if k(scalg)−kLn2(g) = ∞ for all g ∈ [g]. But unfortunately we still cannot prove this. Even if this is true, this alone does not help in our context since we need a complete metric of finite volume with k(scalg)−kLn2(g)<∞which probably cannot be achieved in general.
5. On closed manifolds
The method we used in Theorem 2 for complete manifolds of finite volume allows to reprove the result on closed manifolds with positive Yamabe invariant.
Theorem 14. Let (M, g) be a closed n-dimensional Riemannian manifold with 0 < Q <
Q(Sn). Then, there is a smooth positive solutionv ∈H12 of the Yamabe equation (1).
Proof. The proof in the closed case is essentially the same as the one presented in Section 4.
The only little difference occurs in the proof of Lemma 12 where the cut-off function χR is introduced andQis estimated. We make the following change – we take the smooth cut-off function ηǫ introduced before in Lemma 12. Then with the same estimate as in Lemma 12, where M\Bǫ substitutesB2R andM\B2ǫreplacesBR, we obtain
Q≥ lim
i→∞ Q(B2ǫ, g)(kvikp− kvikLp(M\Bǫ))2−4an
ǫ2 kvikL2(M\Bǫ)+µi
Z
M\B2ǫ
fi2v2idvolg
!
=Q(B2ǫ, g)
For ǫ small enough this gives a contradiction to Q(M)< Q(Sn) due to Lemma 6. Thus, following the rest of the proof in Section 4 we obtain thatv is a smooth positive solution of Lgv=Qvp−1withkvkp= 1. Note that on closed manifolds the conditionk(scalg)−kLn2(g)<
∞of Theorem 2 is trivially fulfilled.
References
[1] Aubin, T.Equations diff´´ erentielles non lin´eaires et probl`eme de Yamabe concernant la courbure scalaire.
J. Math. Pures Appl. (9) 55, 3 (1976), 269–296.
[2] Aviles, P., and McOwen, R. C. Conformal deformation to constant negative scalar curvature on noncompact Riemannian manifolds.J. Differential Geom. 27, 2 (1988), 225–239.
[3] Berard-Bergery, L.Scalar curvature and isometry group. InSpectra of Riemannian Manifolds. Kagai Publications, Tokyo, 1983, pp. 9–28.
[4] Große, N.The spinorial Yamabe equation on complete manifolds of finite volume.in preparation.
[5] Große, N.The Yamabe equation on manifolds of bounded geometry.arXiv: 0912.4398v3.
[6] Hebey, E. Sobolev spaces on Riemannian manifolds, vol. 1635 of Lecture Notes in Mathematics.
Springer-Verlag, Berlin, 1996.
[7] Jin, Z. R.A counterexample to the Yamabe problem for complete noncompact manifolds. InPartial differential equations (Tianjin, 1986), vol. 1306 of Lecture Notes in Math. Springer, Berlin, 1988, pp. 93–101.
[8] Kim, S. Scalar curvature on noncompact complete Riemannian manifolds. Nonlinear Anal. 26, 12 (1996), 1985–1993.
[9] Kim, S.An obstruction to the conformal compactification of Riemannian manifolds.Proc. Amer. Math.
Soc. 128, 6 (2000), 1833–1838.
[10] Schoen, R.Conformal deformation of a Riemannian metric to constant scalar curvature.J. Differential Geom. 20, 2 (1984), 479–495.
[11] Schoen, R., and Yau, S.-T.Conformally flat manifolds, Kleinian groups and scalar curvature.Invent.
Math. 92, 1 (1988), 47–71.
[12] Schoen, R., and Yau, S.-T.Lectures on differential geometry. Conference Proceedings and Lecture Notes in Geometry and Topology, I. International Press, Cambridge, MA, 1994. Lecture notes prepared by Wei Yue Ding, Kung Ching Chang [Gong Qing Zhang], Jia Qing Zhong and Yi Chao Xu, Translated from the Chinese by Ding and S. Y. Cheng, Preface translated from the Chinese by Kaising Tso.
[13] Shubin, M.Essential self-adjointness for semi-bounded magnetic Schr¨odinger operators on non-compact manifolds.J. Funct. Anal. 186, 1 (2001), 92–116.
[14] Trudinger, N. S.Remarks concerning the conformal deformation of Riemannian structures on compact manifolds.Ann. Scuola Norm. Sup. Pisa (3) 22(1968).
[15] Zhang, Q. S.Nonlinear parabolic problems on manifolds, and a nonexistence result for the noncompact Yamabe problem.Electron. Res. Announc. Amer. Math. Soc. 3(1997), 45–51 (electronic).